Download Two-dice horse race - Wiley Online Library

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Probability wikipedia , lookup

Transcript
Original Article
Two-dice horse race
Colin Foster and David Martin
School of Education, University of Nottingham, Nottingham, England
e-mail: [email protected]
[email protected]
Summary
We analyse the “two-dice horse race” task often used in lower secondary school, in
which two ordinary dice are thrown repeatedly and each time the sum of the scores
determines which horse (numbered 1 to 12) moves forwards one space.
Keywords:
Teaching; Statistics; Simulations to estimate probabilities; Two-dice experiment;
Markov processes.
INTRODUCTION
One of us recently saw the classic ‘two-dice
horse race’ lesson in which the teacher asked a
class of Year 8 pupils (age 12–13) to throw two
ordinary dice and add the scores to determine
which horse (numbered 1 to 12) would move
forward one space. The process was repeated
until one of the horses had crossed the finishing
line, which, for a track length of 10, was 10
spaces away.1 The intention was for pupils to
appreciate
simulations and some of which can be framed
in suitable ways for (possibly older) students to
investigate. Intuitively, the probability of horse
7 winning is greater than 16 , but how much
greater and how can this be shown? In general,
what is the chance of a particular horse winning
and how does this depend on the track length?
To calculate these probabilities analytically is
difficult. Below, we consider some aspects of a
two-horse race and then a simulation of the
original race.
1 that horse 1 would never leave the starting gate,
since there is no way to score a total of 1 with
two dice;
2 that horses 2 to 12 are not equally likely to win
and that this must be related to unequal chances
of winning at each throw of the two dice; and
3 that a sample space diagram enables calculation
of the probability of each horse moving one space,
and of a particular horse winning in very
simplified versions of this race.
SIMPLER RACES
In the lesson, the pupils carried out the
experiment with enthusiasm and soon began to
realize that horse number 7 was generally the
best bet. This motivated them to calculate the
probabilities of each horse winning for a track
of length 1 – that is, the probabilities of the various totals when throwing two fair dice – using
the relevant sample space diagram (Foster
2012). However, at this point, some of the
pupils seemed to assume that because the probability of obtaining a 7 with two dice was 16, the
probability that horse 7 would win the race was
also 16 . This raised several possible questions,
many of which could be tackled through
1 For a track length of just 2, Ned can win in two
steps or in three. Students can use a sample space
diagram to show that the chance of Ned winning
is p2 + 2p2q = p2(3 – 2p). Younger students could
do this for selected values of p, and older
students who can factorize simple quadratics
could show that this is greater than p provided
that p > 12.
2 For a track length of 3, ask students how many
steps it could take for Ned to win (3, 4 or 5).
Depending on the class, students could again use
sample space diagrams to find the chance of Ned
winning, either for a specific value of p or
generally in terms of p.
98
Let’s consider two horses, call them Ned and
Pete, with the probability of Ned moving forward
one space each ‘throw’ being p and the corresponding probability for Pete being q = 1 – p.
For some classes, you might wish to use specific
values for p, but make p > 12 . Below are some
activities that could be suitable for different
classes.
© 2016 The Authors
Teaching Statistics © 2016 Teaching Statistics Trust, 38, 3, pp 98–101
Two-dice horse race
99
Table 1. The probability (correct to three decimal places) of each horse winning races of different track lengths
Horse
Track length
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
0.028
0.056
0.083
0.111
0.139
0.167
0.139
0.111
0.083
0.056
0.028
0.009
0.034
0.069
0.113
0.164
0.221
0.164
0.113
0.069
0.034
0.009
0.003
0.021
0.057
0.110
0.179
0.261
0.179
0.110
0.057
0.021
0.003
0.001
0.014
0.047
0.105
0.187
0.293
0.187
0.105
0.047
0.014
0.001
0.001
0.009
0.039
0.099
0.193
0.320
0.193
0.099
0.039
0.009
0.001
0.000
0.006
0.032
0.093
0.197
0.343
0.197
0.094
0.032
0.006
0.000
0.000
0.004
0.027
0.088
0.199
0.363
0.199
0.088
0.027
0.004
0.000
0.000
0.003
0.023
0.083
0.201
0.382
0.201
0.083
0.023
0.003
0.000
0.000
0.002
0.019
0.078
0.201
0.399
0.201
0.078
0.019
0.002
0.000
0.000
0.001
0.016
0.074
0.202
0.415
0.202
0.074
0.016
0.001
0.000
3 This could lead to asking how many steps it could
take for Ned to win for longer track lengths.
Again, the question can be made specific or can
be generalized depending on the class.
4 With three horses on a track length of two,
students could again be asked how many steps it
could take for Ned to win (2, 3 or 4 this time).
Again, if desired, a sample space diagram could
be used to find the chance of Ned winning.
RETURNING TO THE ‘TWO-DICE HORSE RACE’
With our original 11 horses (not counting horse
1), even for a track length of 2, calculating the
probability that any particular horse (say, horse
number 7) will win is complicated, as we have
to consider what all the other horses might be
doing. They will each move 0 or 1 spaces, and,
as
these other
horses
have
different
probabilities from each other of moving, this is
a very messy calculation (an indication is given
in Appendix 1, and a full general approach is
outlined in Appendix 2).
An important question is how many steps (twodice throws) it might take for any particular horse
to win. For a track of length 2, the minimum is 2
and the maximum is 12. The maximum occurs
when all the other 10 horses move one space
and then the winning horse moves two spaces.
Older students could consider the general
question of the number of possible steps for a
horse to win when there are h horses with a track
length of l.2
As the full horse race is clearly far too
complicated for students aged 9–19 to tackle
analytically, there is the opportunity to
demonstrate the usefulness of simulations to
estimate probabilities, with a large number of
simmulations allowing very accurate estimations.
© 2016 The Authors
Teaching Statistics © 2016 Teaching Statistics Trust, 38, 3, pp 98–101
We implemented a simulation3 in which we ran
the horse race 1,000,000 times for different
lengths of track and found by repeating each
simulation a few times that the results were
consistent to three decimal places (Table 1). We
can see from Figure 1 that the distribution
becomes more peaked at horse 7 as the track
length increases.
For students who have met expected values,
there is more of interest. For example, consider
1
horses 6 and 7. We know that p7 – p6 = 36
, so the
expected number of steps of gain of horse 7 over
horse 6 after n throws of the two dice would be
n
36 , which will grow as n grows. We can see that
horse 7’s advantage increases considerably for
longer tracks (Table 1, Figure 1). For the track
length of 10 used in the lesson, the probability
that horse 7 will win is about 0.42, which is much
greater than 16.
Fig. 1. The probability of each horse winning races of
different track lengths
100
CONCLUSION
The explanation for horse 7’s winning
performance in terms of combinations of numbers
that sum to 7 is only a small part of the story; the
length of the track is also important. From a
pedagogical point of view, although using this
task to introduce sample space diagrams might
be helpful, great care in follow-up questions is
needed. To use the two-dice horse race to 10
spaces merely to illustrate the results of a single
throw of two dice may be unhelpful. Simplesounding dice scenarios can be much more
complicated than they might at first sight appear.
Acknowledgements
We would like to thank the editor and the
anonymous reviewers for their very helpful
comments, as well as Philip O’Neill for helping us
with sketching out the details of the Markov
analysis in Appendix 2.
ENDNOTES
1
2
3
An interactive applet for doing this (with a track
length of 12) is available at http://map.
mathshell.org/lesson_support/probability_games/
index.html. An associated lesson plan is available at http://map.mathshell.org.uk/materials/lessons.php?taskid=596.
If the number of horses (not counting horse 1)
is h and the track length is l, then the largest
number of steps to win happens when all horses
have made l – 1 steps, and the winning horse
has made one additional step. Hence, we have
h(l – 1) + 1 goes, which, for an 11-horse race
of track length 2, is 12.
An Excel macro to perform this simulation is
available at http://www.foster77.co.uk/Horse
%20race%20simulation.xlsm The file is not
protected and has extensive comments in the
macro to enable students and teachers to
develop it further if they wish.
Reference
Foster, C. (2012). A straight question.
Mathematics in School, 41(4), 31–34.
APPENDIX 1
We could represent the probability of the ith horse
moving one space forward each time by pi and
Colin Foster and David Martin
then consider the expansion of
12
∑12
. Then,
i¼2 pi
to find the probability that the jth horse will
win the race, we need to sum terms in the expansion in which pj is raised to a higher power
than any other pi and in which the index of
pj ≥ 2. However, the order is important. Imagine
a 3-horse race on a track length of 2, which will
require a maximum of four rounds for one of the
horses to win. The expansion of (p1 + p2 + p3)4
has terms p21p22, for instance, but the order
p1p1p2p2, say, is a win for horse 1, while
p2p2p1p1 is a win for horse 2. So we would need
to add in some of these p21p22 terms but not
others, which introduces a further complexity
to resolve. In general, using combinatorics, we
can say that the probability of horse j winning
our race with a track length of 10 on the nth
throw of the dice will be given by:
P(horse j has moved nine spaces at round n – 1;
all other horses have moved fewer than 10 spaces
at round n – 1). P(horse j moves at round n).
The first of these probabilities is a constrained
sum over the other horses, with multinomial
coefficients to count the sequences which are
the same, so it is not very easy to calculate!
APPENDIX 2
The full procedure is a vector Markov process. Let
Xn = (x2, x3, …, x12), where n is the number of
goes so far (i.e. the number of times the dice have
been rolled) and xk is the space that the kth horse
is on after those n rolls. We start with all horses on
the starting line, so X0 = (0, 0, 0, …, 0). When we
roll the dice, one of the numbers in the vector will
increase by 1. For example, if the first roll gives a
total of 4, then we have X1 = (0, 0, 1, 0, 0, …, 0),
because horse number 4 (confusingly, the third
horse) moves forward one space while all the
others remain where they are. Proceeding in this
way gives us a sequence of vectors X1, X2, X3,
…, which is the vector-valued Markov chain.
Now, the probability that the jth number in our
vector increases by one at each step is the
probability of rolling a total of j, which we are
calling pj, and this gives us the transition
probabilities in the Markov chain; i.e. how likely
it is to move from any particular state to any
other. In this case, almost all of the transition
probabilities between two states are zero, since
from one state, there are only ever 11 other
states that can be reached, corresponding to the
11 possible dice throws. The race finishes as soon
© 2016 The Authors
Teaching Statistics © 2016 Teaching Statistics Trust, 38, 3, pp 98–101
Two-dice horse race
as one of the numbers in the vector reaches 10,
since, at that point, one of the horses has reached
the final square. The standard theory of Markov
chains enables us to work out the probability of
the chain ending up in a particular state (i.e. with
one of the xjs equal to 10 and all the rest less than
10). However, the calculations look prohibitive,
because the state space is enormous (around
1011). If the one-step transition matrix is P, then
© 2016 The Authors
Teaching Statistics © 2016 Teaching Statistics Trust, 38, 3, pp 98–101
101
Pn gives the n-step transition matrix, and, since
the maximum possible number of rounds is 100,
it follows that P100 gives the probabilities needed.
(Note that it is necessary to define carefully
what happens when one horse reaches square
10. The easiest thing is to say that the process
then remains where it is from then onwards, to
make sure that everything works when you get
to n = 100.)